Like everyone else, I am increasingly using AI tools.

Today I finally got around to start setting up what will be the (public) “memory” of my personal future AI agents that will work for me.

Being a developer, I don’t want this to be hidden away in some proprietary black box, but instead want to be able to inspect, edit, and version control it.

I also don’t want it be specific to any one AI tool or provider, but instead be portable and reusable across all of them; whether that’s the (awesome!) Gemini CLI , Anthropic’s Claude Code, OpenAI’s, or my very own Enola.dev .

While project specific instructions belong into AGENTS.md files of the respective projects, such as e.g. my 🐝 Bee project’s AI Agents configuration , I want certain other AI memory that is specific to myself to be shared across all my projects.

As but one small example, when I talk to my AI about “MariaDB4j” it must know that this is my own open source project, who’s canonical GitHub repo is at https://github.com/MariaDB4j/MariaDB4j (and not https://github.com/vorburger/MariaDB4j , which is just my own fork).

So I created https://github.com/enola-dev/vorburger-ai-assistant for this purpose; and it’s git-repos.md to capture the information of the example above.

And with this change , my Gemini CLI installation now automatically pulls this information into my AI agents’ memory. (I’ll later do something similar for other AI tools I use, too.)

This new repo is basically the AI memory equivalent of my Linux dotfiles !

What shall we call such Git repos? How about everyone’s aifiles? 😃